FlowPilot AI Assistant
FlowPilot is Flow-Like’s integrated AI assistant that helps you build, edit, and understand your workflow automations using natural language. It supports multiple AI providers and specialized agents for different tasks.
Overview
Section titled “Overview”FlowPilot can:
- Create nodes and connections from natural language descriptions
- Explain existing flows and help you understand complex automations
- Debug issues by analyzing execution logs and suggesting fixes
- Search the node catalog to find the right nodes for your task
- Generate complete workflows from high-level requirements
AI Providers
Section titled “AI Providers”FlowPilot supports two AI provider modes:
Bits (Default)
Section titled “Bits (Default)”Uses your configured model bits from your user profile. This is the default mode and works with any LLM provider you’ve set up in Flow-Like.
GitHub Copilot
Section titled “GitHub Copilot”Uses the GitHub Copilot SDK directly for AI-powered assistance. This mode provides access to GitHub’s latest AI models and features native integration with your GitHub account.
Setting Up GitHub Copilot
Section titled “Setting Up GitHub Copilot”To use FlowPilot with GitHub Copilot, you need to install and configure the GitHub Copilot CLI.
Prerequisites
Section titled “Prerequisites”- An active GitHub Copilot subscription
- One of the supported platforms:
- macOS
- Linux
- Windows (PowerShell v6 or higher)
Installation
Section titled “Installation”Choose your preferred installation method:
macOS and Linux (Homebrew)
Section titled “macOS and Linux (Homebrew)”brew install copilot-cliFor the prerelease version:
brew install copilot-cli@prereleaseWindows (WinGet)
Section titled “Windows (WinGet)”winget install GitHub.CopilotFor the prerelease version:
winget install GitHub.Copilot.Prereleasenpm (All Platforms)
Section titled “npm (All Platforms)”npm install -g @github/copilotFor the prerelease version:
npm install -g @github/copilot@prereleaseInstall Script (macOS and Linux)
Section titled “Install Script (macOS and Linux)”curl -fsSL https://gh.io/copilot-install | bashOr with wget:
wget -qO- https://gh.io/copilot-install | bashAuthentication
Section titled “Authentication”After installation, you need to authenticate with your GitHub account:
- Launch the CLI by running
copilotin your terminal - Follow the login prompt - enter the
/logincommand if prompted - Complete the authentication in your browser
Alternative: Personal Access Token (PAT)
Section titled “Alternative: Personal Access Token (PAT)”You can also authenticate using a fine-grained PAT:
- Visit GitHub Token Settings
- Under “Permissions,” add the “Copilot Requests” permission
- Generate your token
- Set the
GH_TOKENorGITHUB_TOKENenvironment variable with your token
Verifying Installation
Section titled “Verifying Installation”Run copilot in your terminal. You should see the Copilot CLI interface with authentication status.
Using FlowPilot
Section titled “Using FlowPilot”Switching Providers
Section titled “Switching Providers”In the FlowPilot panel, you’ll see provider toggle buttons:
- Bits - Use your configured model bits
- Copilot - Use GitHub Copilot
Click on the Copilot button to switch. In the desktop app, FlowPilot will automatically connect to the local Copilot CLI. In the web version, you’ll be prompted to enter a server address.
Selecting Models
Section titled “Selecting Models”When using GitHub Copilot, you can choose from available models:
- Claude Sonnet 4.5 (default)
- Claude Sonnet 4
- GPT-5
- And other available models
Use the model selector dropdown to switch between models.
Chat Interface
Section titled “Chat Interface”The FlowPilot chat interface allows you to:
- Ask questions about your current flow
- Request modifications to nodes and connections
- Get explanations of what specific nodes do
- Debug errors by sharing log output
Example prompts:
- “Add an HTTP request node that calls the OpenAI API”
- “Connect the output of the JSON parser to the email sender”
- “Explain what this flow does”
- “Why is this node failing? Here’s the error…”
Specialized Agents
Section titled “Specialized Agents”FlowPilot includes specialized agents optimized for different tasks:
Frontend Agent
Section titled “Frontend Agent”Focused on UI/UX development:
- Creating responsive UI components
- React patterns and hooks
- CSS/Tailwind styling
- A2UI component system
Backend Agent
Section titled “Backend Agent”Focused on workflow and data processing:
- Flow graph design and node connections
- Data transformation and processing
- API integrations
- Error handling and retry logic
General Agent
Section titled “General Agent”Can handle both frontend and backend tasks, automatically determining the best approach based on your request.
How It Works
Section titled “How It Works”When you send a message to FlowPilot, it:
- Analyzes your request and the current board context
- Searches the node catalog to find relevant nodes
- Plans the changes needed to fulfill your request
- Generates commands to modify the board
- Applies changes while maintaining undo/redo history
All changes are applied through the same command system used by manual editing, so you can always undo AI-generated changes.
Infinite Context
Section titled “Infinite Context”FlowPilot with GitHub Copilot supports infinite context handling, which automatically manages conversation length by:
- Compacting older messages when the context window fills up
- Preserving important context while discarding less relevant details
- Allowing long-running sessions without losing track of the conversation
This is particularly useful for complex workflows that require multiple iterations to build.
Privacy and Security
Section titled “Privacy and Security”- Local processing: The Copilot CLI runs locally on your machine
- GitHub authentication: Uses your existing GitHub credentials
- No data storage: Conversations are not stored on GitHub’s servers beyond what’s needed for processing
- Explicit approval: All changes require your approval before being applied
Troubleshooting
Section titled “Troubleshooting”Copilot CLI not found
Section titled “Copilot CLI not found”Ensure the Copilot CLI is installed and in your PATH. Try running copilot directly in your terminal.
Authentication issues
Section titled “Authentication issues”Run copilot in your terminal and use the /login command to re-authenticate.
Connection issues (web version)
Section titled “Connection issues (web version)”Ensure your Copilot server is running and accessible from your browser. Check firewall settings if needed.
Model not available
Section titled “Model not available”Some models may require specific Copilot subscription tiers. Check your subscription level if a model is unavailable.